УДК 621.396.21 O. Pavlov MATRIX FORM OF THE LSP-HS METHOD National Technical University of Ukraine 'Kyiv Polytechnic Institute' Abstract The coding theory of the forms of spectral envelope of speech signals of the using linear spectral parameters of the highest splitting (LSP-HS) is regarded. Advantages of using of LSP-HS parameters compare with classic LSP parameters are specified. It is demonstrated that the direct and inverse transform of LSP-HS method can be regarded as a certain matrices transform from the LPC coefficients. The formula of estimates of the spectral envelope of the synthesized speech signal in space LSP-HS is written. Keywords: Encoding shape of the spectral envelope, linear spectral parameters (LSP), linear spectral pairs (projection LSPr), linear spectral frequencies (LSF), linear prediction coefficients (LPC). Introduction The new original method of encoding of spectral envelope of speech signal (that had been named as LSP-HS — linear spectral parameter of the highest splitting) has been described in previous works [1 — 7]. This method is recommended to using in the speech transformation devices of the receiving and transmitting equipments, which are based on the linear prediction (LP) algorithms. The main idea of method is in that characteristic polynomial A(z) of prediction filter of M-orded, which represented as one stable polynomial of M-order, ∑ ∑ = - = - + = ′ - = M i i i M i i i z a z a z A 1 1 1 1 ) ( , (1) is proposed to be represented as M elementary stable normalized polynomials of 1th order 1 1 1 ) ( - + = z a z A vvvv vvvv , (2) which are the results of step-by-step splitting of original polynomial A(z), fig. 1. The roots of elementary polynomials (2), 1 1 1 ) ( - + = z a z A vvvv vvvv , are the linear spectral projections of the highest splitting (LSP-HS). Arccosine from roots of elementary polynomials (2), 1 1 1 ) ( - + = z a z A vvvv vvvv , are the linear spectral frequencies of the highest splitting (LSF-HS).